hoploop/chainer repository overview

⁠QUALIA IT Personal Assistant

Artificial Intelligence and Reactive Event-based python systems join together.

⁠Version 4.5.0 contains several changes, including:

• Centralized process oriented ZeroMQ, including common shared Threadpool, Logging and ZeroMQ Socket Context
• New action and reaction engine, including action configuration in the frontend

⁠Main features

• Image Classification
• Object Detection
• Face Recognition
• Natural Language Sentiment Analysis
• Chatbot
• Every day utilities: notes, weather
• Action reaction qualias mechanism

⁠Engine Main Features

• Asynchronous calls are the main approach, used as much as possible
• Event-based programming paradigm through pure Publisher/Subscriber architecture
• Publisher/Subscriber ZeroMQ with topic filtering
• Possibility to Balance and scale nodes with multiple Publishers and Subscribers
• Messagepack json compression
• Python Dataclasses standard object representation
• Websockets
• Document driven database MongoDB
• Json Type Serializer used at Typescript level
• Dockerizable with docker-compose

⁠Frontend main functionalities

• Microphone speech detection
• Microphone sound recording
• Speaker text-to-speech
• Webcamera image processing
• Geolocation
• Responsive mobile-first bootstrap based
• Angular 9.x
• Multilanguage support (en,it)

⁠Integrations with external systems:

• Netatmo
• Openweather
• Wikipedia
• JiraServer Rest api

⁠How to use it

⁠Requirements

Main requirements are a machine with:

• Docker installed (docker mode)
• Python >= 3.7.x installed (dev mode)
• Web Browser (Chrome, Safari, Firefox supported)
• OpenSSL installed for generating the certificates (initial setup)
• NodeJS installed (dev mode)

⁠Running the full system in dev mode with the compact run

1. Generating the ZeroMQ Certificates is done by: common/setup.sh
2. Run the MongoDB: docker-compose up mongo or have your local mongo running and available on loopback 127.0.0.1
3. Run the all the system with ZeroMQ Inproc: ./start.sh

⁠Running separate services in dev mode

1. Generating the ZeroMQ Certificates is done by: common/setup.sh
2. Run the MongoDB: docker-compose up mongo or have your local mongo running and available on loopback 127.0.0.1
3. Run the chainer service: chainer/start.sh
4. Run the mouth service: mouth/start.sh
5. Run the hear service: hear/start.sh
6. Run the system service: system/start.sh
7. Run the sight service: sse/start.sh
8. Run the plugins service: plugins/start.sh
9. Run the sse service: sse/start.sh
10. Finally run the frontend: sse/start.sh
11. Go to http://127.0.0.1:4200 with your favourite browser

⁠Running the full system in production ready mode

The system is governed by a full Docker system through docker compose. Before starting the application, you should generate the certificates.

1. Generating the ZeroMQ Certificates is done by:
   1. Go to common folder and: common/setup.sh
   2. Go to services/nginx folder services/nginx/setup.sh
2. Build the frontend web/build.sh

Execute docker-compose up and go to your local browser at https://127.0.0.1, the docker compose will launch

• Sight docker image (image parsing and classification, face and object recognition)

• Mouth docker image (output message and voice)

• Hear docker image (hear sense, text parsing and sentiment analysis)

• System docker image (main data storage handler and authentication)

• Plugins docker image (plugin handlers)

• SSE docker image (websocket forwarder)

• Chainer docker image (qualias action reaction engine)

• Mongodb docker image (data persistence service)

• NGINX docker image (web reverse proxy handler)

• Portainer docker image (docker services handler)

• Socat docker image (docker services handler helper)

• Maxtext docker image (sentiment analysis for text)

⁠Technical standards

1. Semantic versioning for git and npm: [major].[minor].[patch]-[build/beta/rc]
2. Angular 9.x with Ivy AOT Compiler
3. Standard Python logging formatting with rotation

⁠Local installation of OpenSSL on MACOSX1

In order to generate the certificates, it is required to have the openssl installed in your system

1. Install homebrew /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install.sh)"
2. Install openssl with brew: brew install openssl
3. Install tesseract with brew: brew install tesseract this is required for OCR Pytesseract to work
4. Install the sound file encoder: brew install libsndfile

⁠Local installation of Kubernetes on MacOSX

1. Install homebrew /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install.sh)"
2. Install minikube brew install minikube
3. Install kompose
   1. Download the binary curl -L https://github.com/kubernetes/kompose/releases/download/v1.21.0/kompose-darwin-amd64 -o kompose
   2. Make it executable chmod +x kompose
   3. Move it to the bin folder to be available on the system sudo mv ./kompose /usr/local/bin/kompose
4. Start minikube minikube start
5. Start the docker on minikube kompose --file docker-compose.yml up

Optional steps are:

• Enable metrics-server minikube addons enable metrics-server
• See metrics kubectl get apiservices

⁠Next steps

• Evaluating the possibility to use a generic json schema based generator, with proper serializers and deserializers, marshallers and unmarshallers